[4M3-GS-13] AI application: Electric power

[4M3-GS-13-05] Development of Prediction Method for Household Power Consumption

〇Soehi Suzuki¹, Naoya Takeishi², Masanari Kishimoto¹, Naohisa Yoshitani¹, Kenta Okino¹, Yoshinobu Kawahara², Masahito Hiroshima¹ (1. KYOCERA Corp., 2. RIKEN Center for Advanced Intelligence Project)

Keywords:Time Series Prediction, Power Consumption, Virtual Power Plant

Since the Great East Japan Earthquake of 2011, the vulnerability of the centralized energy system has surfaced, and the system is being reviewed. In particular, a shift to a distributed energy system that utilizes renewable energy has been demanded due to consideration for the environment, and a concept called VPP (Virtual Power Plant) has attracted attention as one solution. For the supply-demand adjustment market established in 2021, the aim of this study is to develop a highly accurate prediction method of power consumption in household. In the prediction, it is important to capture fluctuations caused by the rhythm of the residents' lives. In this paper, we propose an autoregressive model that considers the weight of data samples. The effectiveness of the proposed method was confirmed using 30-minute granularity time-series data obtained from multiple households.